In order to interconnect interior devices of an apparatus or to interconnect different apparatus, as a conventional construction, Japanese Laid-open Patent Application Nos. 2000-164667 and 2000-164666 disclose to interconnect them through a standard LAN (local area network), such as Ethernet (registered trademark).
As another known example, Japanese Laid-open Patent Application No. 9-153441 divides a LAN into a plurality of segments and installs a processing station between the divided segments to copy data.
Japanese Laid-open Patent Application No. 11-85326 discloses a system having a plurality of computers interconnected through a network, and all the design information is previously transferred from the client to a plurality of servers.
Further, Japanese Laid-open Patent Application No. 2002-132986 discloses a system which interconnects clients and a manufacturing apparatus using the Internet.
An electron beam lithography apparatus is disclosed in Japanese Laid-open Patent Application No. 63-208215, wherein a plurality of electron beam lithography systems are respectively connected with a buffer memory for storing writing data, and a control computer controls these plurality of buffer memories such that desired image data is stored in each buffer memory from the writing data storing unit, thereby continuously writing different patterns within a writing area of each electron beam lithography system. Japanese laid-open Patent Application No. 7-307262 discloses an electron beam lithography apparatus which draws desired patterns by a charged electron beam with the aid of apertures and the like based on CAD data as semiconductor design information.
As to conventional storage area networks, WO00/18049 and WO00/17769 disclose a link through a fiber channel. WO00/29954 discloses a network through an optical fiber. Also, a link through Ethernet (registered trademark), such as iSCSI, iFCP, and FCIP, and a link through a switched bus or a shared bus are known. The storage area network is a general term of the network for linking storage devices without consideration of a kind of communication device. The link of storage devices through a serial bus as defined in IEEE1394 and the link of storage devices through a switched bus as defined by InfiniBand (registered trademark) are also included in the storage area network.
Mask layout data as a kind of semiconductor design information is prepared by a logic design maker. The mask layout data is then processed by the semiconductor design apparatus to provide a mask (reticle). The mask layout data is stored in a local storage device of the logic design maker. If the logic design maker has to supply the mask layout data, for example, to a mask shop which possesses a semiconductor manufacturing apparatus, the mask layout data should be copied in a storage medium such as a magnetic tape. The mask shop then receives the storage medium and copies the contents of the storage medium into a local storage device of the mask shop.
However, the aforementioned conventional technologies do not consider the kind of data flowing through the network. Because two kinds of data, i.e. a large volume of CAD data representing design information of semiconductors and message data representing control commands for controlling and linking a variety of devices, are transferred through the same network, the traffic inevitably increases, degrading the performance of the network, which in turn adversely affects the overall performance of the system. In other words, the conventional networks have a drawback in that the throughput of the network changes according to the frequency of issuing the control command, the frequency of generating a response to the command, and the transmission/reception of a large volume of data, thereby degrading the overall performance of the apparatus. As the advance of the micro-fabrication technology in particular, the volume of the design data of semiconductors and masks and the volume of the image data as the inspection result drastically increase. As a result, the band of the network is occupied by simply communicating these data. This adversely affects the transmission and reception of the message data.
As a prior art technology to solve this problem, all the design information is previously transferred to a plurality of computers for processing. However, because the volume of data transfer increases as the number of computers linked, extreme amount of traffic occurs at time of the data transfer. Further, each of the plurality of computers for receiving the design information must provide a storage device for storing a large volume of design information.
In this prior art technology, CAD data that is the basis of the design information of semiconductors is converted into a writing data format originated from the electron beam lithography apparatus, and the pattern data indicated by this writing data format is further processed such as by conversion and correction in real time operation, thereby radiating an electron beam. These processes are sequentially and continuously executed. Therefore, the conversion process and the correction process are carried out independently before executing the writing, and it is impossible to temporarily store the processing results. As a result, it is very difficult to predict the time required for electron beam radiation and the accuracy of writing. Because processing results can not be stored in mid-course of the operation, it is very difficult to suspend and restart the process. Even in the case of processing the same design data, the conversion process and the correction process must be repeated from the beginning.
In these prior art technologies, data is mostly stored in a file system which realizes data having arbitrary length as assemblies of a plurality of blocks having fixed length. This file system has a control list indicating the relation of a plurality of fixed length blocks associated with the arbitrary data. However, a large volume of fixed length blocks are required against such a large volume of data, which leads to a large volume of the control list. This decreases an area in which the storage device actually stores data, and also adversely deteriorates the throughput because of the retrieval process of the control list for accessing the data. The fixed length blocks are ineffectively arranged in the storage device as the result of preparation, deletion or transfer of the data, which also deteriorates the throughput.
Of the above prior art technologies, a technique is suggested wherein a LAN is divided into a plurality of segments and processing stations are installed between the segments to perform copying of the data for the purpose of alleviating the traffic. However, because the processing stations copy data between the segments, the processing stations per se become a bottleneck of the overall performance of the system. Further, because each of the storage devices interconnected to individual segments copies the same data, the consistency management of the copied data becomes complicated, which results in difficulty in system operation. For example, even if the semiconductor inspection apparatus and the semiconductor manufacturing apparatus are interconnected through the network, data must be copied through the network in order to transfer the data between these apparatus. This results in a crowd of the network and deteriorated throughput. Even in the case where a plurality of semiconductor inspection apparatus and a plurality of semiconductor manufacturing apparatus are interconnected through the network and processing is carried out in a parallel manner, data must be copied through the network. This also results in a crowd of the network and difficulty in the system organization due to management of data exchange. Further, in most cases, it is impossible to interconnect a new storage device through the network without stopping the operation of the system. In other words, when the storage device is filled up, it is very difficult to extend the storage capacity.